Development of a digital twin for predicting burr height during axial drilling

Burr formation is a major concern in the drilling of aeronautical components, as it negatively affects assembly quality and fatigue life. In the context of Industry 4.0, digital twins are expected to enable the prediction of burr formation and the active control of drilling processes. However, this requires robust predictive models, for which the selection of appropriate input variables and signal features is a fundamental step. In this work, axial force and torque signals acquired during Ti6Al4V drilling with the same cutting tool under four tool conditions were analysed, and after machining, the corresponding burr height at the hole exit was measured. The relevant descriptors were identified using a signal-processing methodology that extracts temporal and spectral domain features related to signal amplitude, energy, power, and dynamic behaviour. The results show that axial force-based descriptors extracted from the stable phase of drilling exhibit a relationship with burr height, following power-law trends with coefficients of determination up to R-squared higher than 0.83, although this behaviour is largely governed by the separation between wear levels. Comparable performance was obtained using the force at tool breakthrough, highlighting the potential for developing a digital twin to anticipate burr formation before the exit phase, an important capability for future real-time monitoring and control strategies.